Answers are grounded in this paper and cite their sources.
Interior permanent magnet (IPM) motors are widely used in electric vehicles (EVs). Therefore, it is desirable to design motor components with various shapes in a rapid and automated manner. However, traditional design processes rely heavily on trial-and-error, making them time-consuming. This paper proposes a deep learning-based methodology for generating rotor shapes that satisfy specified conditions, namely, magnet area and torque. To improve accuracy, we incorporate a physics-guided generative adversarial network (PG-GAN), which distinguishes realistic designs from unrealistic ones based on the discrepancy between the desired and actual performance metrics. These performance metrics are computed using JMAG, a general-purpose electromagnetic simulation tool. After training, the PG-GAN produces physically consistent and feasible designs. In this study, we use a VAE/WGAN-gp model as a pretrained base and formulate a physics-guided extension of it. The results show that the proposed PG-GAN significantly outperforms both the standard GAN and the conventional VAE/GAN in generating accurate rotor shapes. • IPM motor rotor part is generated using a generative model. • Physics-guided VAE/WGAN-gp model is proposed to improve accuracy. • The trained model generated reasonable motor rotor parts in high accuracy.
No takes yet. Share an insight, caveat, or question.
Yonekura et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: